Setting up the Atlassian Rovo MCP Server
Connect Atlassian's official MCP server to Claude Code, Codex, and others via OAuth or API token — with tightly scoped permission groups.
- Skill Road
- Setting up the Atlassian Rovo MCP Server
Published on 09.09.2026
The Atlassian Rovo MCP Server is hosted: nothing is installed locally, and the MCP client connects directly to mcp.atlassian.com. Atlassian currently recommends version 2 exclusively, at https://mcp.atlassian.com/v2/mcp.
Setting it up in Claude Code
claude mcp add --transport http atlassian https://mcp.atlassian.com/v2/mcp
Then run /mcp in a session to start the OAuth sign-in. After confirming in the browser, the Atlassian tools are ready.
Setting it up in Codex
codex mcp add atlassian --url https://mcp.atlassian.com/v2/mcp
Setting it up in other clients
For Cursor, VS Code with GitHub Copilot, ChatGPT, Gemini CLI, and Windsurf, the official repository provides ready-made install links or marketplace entries. Clients without native remote support connect via the mcp-remote npm package and the server URL https://mcp.atlassian.com/v2/mcp.
Authenticating without an interactive login
For automations without browser access, a personal API token (basic auth) or a service account API key (bearer token) can be used instead of OAuth. Both must first be enabled by an organization admin under Atlassian Administration → Rovo → Rovo MCP server → Authentication. For Jira Service Management tools, API token authentication is the only supported option anyway.
Keeping access tight
The tool catalog is organized into per-product permission groups (read, write, search, delete, manage). Organization admins should initially grant newly connected clients read-only access and only enable write, delete, or manage permissions once the agent proves reliable. Every action is still limited to whatever the signed-in account can already see and do in Atlassian Cloud.
Moving from v1 to v2
Per Atlassian, existing v1 connections automatically start using v2 tools. The legacy SSE endpoint https://mcp.atlassian.com/v1/sse will be retired after 30 June 2026; anyone with a client still configured for it should switch to /v2/mcp in time and may need to clear cached client IDs or .well-known credentials if authentication fails afterward.
Understanding permission groups in detail
Since the tool catalog is split into per-product permission groups, it's worth checking the specific mapping before granting access: a tool for creating Confluence pages falls into a different group than one for deleting Jira tickets, even though both touch the same Atlassian workspace. Organization admins should enable these groups per product rather than blanket-granting all of them.
Verifying the setup
After OAuth confirmation, test with a harmless, read-only task, such as "Show me the latest Jira tickets in project X" or "Search Confluence for onboarding documents." Only enable additional permission groups once this works reliably.
This staged rollout costs a bit of setup time but prevents an insufficiently tested agent from accidentally making far-reaching changes to production Jira or Confluence data.
Source: github.com/atlassian/atlassian-mcp-server and support.atlassian.com/atlassian-rovo-mcp-server, checked on 2026-09-06.
Frequently asked questions
Do I need to install anything for the Atlassian MCP Server?
No. The server is hosted by Atlassian and connected via a URL. Only clients without native remote support start a small bridging process through the mcp-remote npm package; the server itself is not installed.
What does the Atlassian MCP Server cost?
The server itself can be used without a separate fee. Cost only comes from your existing Atlassian plan (Jira, Confluence, etc.), not from the MCP access.
Should I use v1 or v2?
For all new setups, v2 at https://mcp.atlassian.com/v2/mcp. The legacy SSE endpoint from v1 will be retired after 30 June 2026; per Atlassian, existing v1 connections automatically start using v2 tools.
How do I stop an agent from accidentally deleting issues?
Delete and manage are separate permission groups that an organization admin must enable specifically. As long as only read and, if needed, write are enabled, the agent cannot delete anything.
Can I use the server without a browser login in a pipeline?
Yes. Instead of OAuth, you can use a personal API token or a service account API key, which an organization admin must first enable. For Jira Service Management, that is the only supported option anyway.